The Broken Windows Theory of Manufacturing
How recurring abnormalities become normal — and what they reveal about the people, decisions and culture behind a factory's performance.
In this article
By Anas BOURAS | Founder, ALFABE
Imagine a new operator joining an automated production line.
During a transfer cycle, a low-vacuum warning appears on the machine's display. The cycle finishes, the warning disappears, and the operator turns to an experienced colleague.
“Don't worry. It always does that.”
This is a hypothetical example, but the conversation illustrates something worth examining in manufacturing.
Perhaps the warning has already been investigated. Engineering understands its cause, the operating conditions have been assessed, and the team knows exactly how to respond.
Or perhaps the warning has appeared so often that nobody considers it unusual anymore. The equipment keeps running, production targets are met, and the original reason for the warning has gradually disappeared from everyone's attention.
Now imagine that same warning appearing for the first time.
Someone would probably ask what happened. Maintenance might be called, the equipment checked, and the condition recorded.
A few weeks later, the response could be very different.
The warning hasn't necessarily changed. The way people interpret it has.
And that's what interests me. A factory's standards tell us what should happen. Its daily decisions show us what people have learned to accept.
The idea behind the broken window
In 1982, James Q. Wilson and George L. Kelling published Broken Windows: The Police and Neighborhood Safety.
Their argument was that visible signs of disorder, when left unattended, could influence people's expectations about acceptable behaviour and whether others would intervene.
The theory became influential in discussions about crime and policing, although its broader causal claims remain contested. A 2018 review by the National Academies of Sciences, Engineering, and Medicine found mixed evidence across different broken-windows policing approaches.
I wouldn't claim that this theory proves how a factory behaves. An unattended machine warning is not equivalent to a broken window in a neighbourhood, and correcting minor housekeeping issues doesn't automatically prevent industrial accidents.
But the analogy raises an interesting question: can repeated exposure to an abnormal condition change what people consider normal?
In manufacturing, a more directly relevant concept is normalization of deviance.
Sociologist Diane Vaughan explored this phenomenon in her research into the Challenger disaster. Her work examined how departures from expected technical conditions could gradually become accepted through organizational decisions, particularly when previous occurrences hadn't produced the consequences people feared.
The Columbia Accident Investigation Board later examined related organizational problems, including how earlier successful outcomes influenced the interpretation of technical risk.
These were catastrophic events, and their consequences shouldn't be casually compared with routine production issues. What makes the research relevant is the way organizations interpret evidence.
When something unexpected happens repeatedly without an obvious failure, people may become more confident that the condition is acceptable.
But the fact that a process has continued operating doesn't establish how much safety or process margin remains, whether the underlying mechanism is understood or what will happen when conditions change.
Experience tells us what has happened. Engineering must also establish what could happen next.
When the warning stops being a warning
Let's follow our hypothetical vacuum-handling station.
The machine has an early-warning threshold intended to indicate deteriorating vacuum performance before the applicable transfer-inhibit limit is reached. Its protective function is designed and verified to prevent transfers when the required holding conditions cannot be established.
At first, an intermittent warning is reported and investigated. Maintenance restores operation, but the underlying cause remains uncertain.
The warning returns several days later. The team checks the equipment again and production resumes.
Eventually, the response becomes familiar. Operators recognize the message, technicians know the recovery procedure, and supervisors have become accustomed to the occasional interruption.
Nobody has formally approved a different operating standard. Yet the warning may now attract less attention than it did when it first appeared.
That doesn't automatically mean the factory has lost control. A recurring condition might be known, assessed and managed under authorized temporary controls.
The distinction lies in what the organization actually knows and does.
Suppose a more detailed investigation eventually identifies a deteriorating vacuum hose. Before deciding whether production can continue, engineering needs to understand the impact on holding capability, verify that the monitoring and protective functions remain effective, and establish the relevant operating limits.
Quality also needs to consider the product handled during the period of uncertainty.
If there is reason to believe that required handling conditions may not have been maintained, the team must identify potentially affected production, assess the available evidence and apply appropriate containment and product disposition. Completing a transfer without visibly dropping a component is not, by itself, sufficient evidence of conformity.
If continued safe and compliant operation cannot be demonstrated, the equipment must not continue operating merely because previous cycles were completed successfully.
Those decisions are determined by the actual technical risk.
But there's another question I would want answered.
Why did it take several occurrences for the organization to reach this point?
Perhaps the early investigations lacked enough evidence. Perhaps the fault was intermittent and difficult to reproduce. Perhaps maintenance had identified the probable cause but needed an approved replacement component or a shutdown window.
All of those explanations are plausible.
What concerns me is a different situation: the problem keeps returning, but the investigation stops progressing because everyone has become comfortable with the recovery routine.
The team may be very good at restoring production while the original cause remains unresolved.
That's the point at which familiarity can begin replacing technical evidence.
The gap between the standard and daily practice
Every factory has defined requirements: process specifications, maintenance instructions, control plans, reaction plans and operating procedures.
But spend enough time on the shop floor and you'll also discover what people routinely accept in practice.
I think of the distance between those two things as a factory's tolerance boundary.
It's not a formal engineering measure or an alternative to risk assessment. It's a way of describing the gap that can develop between what the organization requires and what its daily decisions allow.
That boundary rarely changes through one dramatic announcement.
It can move through a succession of small decisions.
The first warning triggers an investigation. The second is handled through a familiar intervention. After several occurrences, the interruption is expected. Eventually, someone new joins the team and is told that the machine has always behaved that way.
The condition may be physically unchanged, but the organization's expectations have shifted.
That is why I find recurring workarounds interesting during factory visits.
A temporary repair might be technically justified and properly controlled. A repeated maintenance intervention might be part of a planned strategy. Some problems remain open because permanent correction requires investment, engineering work or unavailable resources.
Those situations should not automatically be interpreted as a lack of discipline.
What matters is whether the decisions behind them are still understood.
If an abnormal condition remains in place, can the team explain its technical significance, who authorized the current approach, what is being monitored and what would trigger reassessment?
If the answer is clear and supported by evidence, the organization may be managing a legitimate constraint.
If the answer is simply that things have always been done that way, I'd want to investigate further.
Visual management and 5S can help make deviations apparent, but they cannot substitute for the response that follows. A warning displayed for months without meaningful follow-up eventually risks becoming part of the background.
The problem is no longer a lack of visibility.
It's that visibility has stopped producing useful action.
When the problem becomes a management decision
Return to the deteriorating vacuum hose.
Suppose engineering has established the cause and the equipment supplier confirms that an approved replacement component will take several days to arrive. The intervention also requires a planned shutdown of a bottleneck machine.
Production has customer commitments. Maintenance needs access to the equipment. Procurement is trying to accelerate delivery.
At this point, the technical diagnosis may be straightforward while the decision about how to proceed is not.
The first responsibility is to establish which options are permissible. If continued operation cannot meet the applicable safety, process or product requirements, delivery pressure doesn't change that conclusion.
If a limited period of operation is technically justified and permitted under approved temporary controls, the responsible functions can establish the conditions, monitoring, escalation criteria and deadline for permanent correction.
Those decisions should be made by people with the appropriate technical competence and authority. An effective organization doesn't need every maintenance issue to reach the plant director, but it does need clear rules for when a problem must be escalated.
There's also a business dimension that I think is frequently underestimated.
A recurring interruption might appear inexpensive when considered individually. Across several weeks, however, it may consume technician hours, engineering time, production capacity and additional quality checks.
Those costs are usually distributed across departments.
The permanent correction may appear expensive in one budget, while the cost of repeatedly recovering the machine is absorbed elsewhere.
That can make postponing the correction look more reasonable than it really is.
A competent maintenance team might restore production quickly every time. That's valuable. But management still has to determine whether continuing that recovery strategy is justified.
Once the replacement is completed, the problem isn't closed simply because the work order has been signed.
Engineering needs evidence that the original vacuum performance has been restored, the required controls function correctly and the warning no longer recurs under relevant operating conditions. Quality must also complete any outstanding assessment and disposition of potentially affected product.
The corrective action should be closed because its effectiveness has been verified.
Not because the machine is running again.
What the operator learns
Now return to the person who first noticed the warning.
Imagine that the operator reported the abnormality several times. Maintenance responded, but the permanent repair was delayed while the replacement part was arranged.
If nobody explained what was happening, the operator might reasonably wonder whether reporting it again served any purpose.
The warning continued to appear. Production continued. The response appeared to be the same.
Eventually, the operator might stop mentioning it.
A supervisor could interpret that as a failure to follow the escalation procedure. But I'd also want to understand what happened after the previous reports.
Did anyone explain the investigation? Was the operator told which conditions required an immediate stop? Did the team understand why a temporary operating arrangement had been authorized, assuming one was permissible?
Or did they simply see the same warning every day while everyone carried on?
The difference matters because people learn how an organization works from what happens when they raise a concern.
If reporting an abnormality leads to a clear investigation and useful feedback, the process gains credibility.
If reporting the same problem repeatedly produces no visible response, some people may gradually stop expecting anything to change.
That can affect the information management receives.
A reduction in reported abnormalities might reflect better equipment performance. But it could also reflect a change in reporting behaviour while the underlying condition remains.
A dashboard alone cannot distinguish between those explanations.
We need to look at equipment data, reporting history, the actions taken and what people are experiencing on the shop floor.
And we should be careful before turning an organizational problem into an individual discipline problem.
Operators are responsible for following defined requirements. Technicians and engineers have responsibilities too. But those requirements need to be understood, practically achievable and supported by the people making operational decisions.
Sometimes a recurring workaround reveals a technical weakness. Sometimes it reveals a lack of resources or an ineffective escalation process. And sometimes the organization has made the expected behaviour so frustrating that people have gradually stopped attempting it.
Correcting the behaviour without understanding its cause may leave the real problem untouched.
What have we stopped noticing?
The Broken Windows analogy doesn't lead me to believe that every imperfection in a factory is a sign of poor management.
Machines deteriorate. Components fail. Processes vary. Temporary solutions are sometimes necessary, and not every deviation deserves the same level of attention.
The most visible abnormality isn't necessarily the most dangerous.
What interests me is how an organization decides which conditions are acceptable, how those decisions are communicated and whether people continue to question the assumptions behind them.
During a factory walk, I would still examine the technical condition, the maintenance history and the available process data.
But I would also listen carefully to the explanation given by the people working with the equipment.
When someone says, “It always does that,” they might be describing a well-understood condition that has been properly assessed and controlled.
Or they might be describing a problem that has existed for so long that nobody expects it to be fixed anymore.
The difference isn't always visible from the outside.
You discover it by understanding what happened when the abnormality first appeared, what decisions followed and why the current situation remains in place.
A factory can become very effective at recovering from recurring problems without necessarily becoming better at eliminating their causes.
And when the same condition persists long enough, it may stop appearing in daily conversations even though it hasn't disappeared from the process.
That's the manufacturing lesson I take from the Broken Windows idea.
What have we stopped noticing?
Anas BOURAS Founder, ALFABE
References and further reading
- Wilson, J. Q., and Kelling, G. L. (1982). Broken Windows: The Police and Neighborhood Safety. The Atlantic. https://www.theatlantic.com/magazine/archive/1982/03/broken-windows/304465/
- National Academies of Sciences, Engineering, and Medicine (2018). Proactive Policing: Effects on Crime and Communities, Chapter 4. https://www.nationalacademies.org/read/24928/chapter/6
- Vaughan, D. (1996). The Challenger Launch Decision: Risky Technology, Culture, and Deviance at NASA. University of Chicago Press.
- Columbia Accident Investigation Board (2003). Columbia Accident Investigation Board Report, Volume I, Chapter 8: History as Cause — Columbia and Challenger. https://sma.nasa.gov/SignificantIncidents/assets/columbia-accident-investigation-board-report-volume-1.pdf
- Lean Enterprise Institute (2014). Reinforcing Lean Behavior Through Visual Management. https://www.lean.org/the-lean-post/articles/reinforcing-lean-behavior-through-visual-management/
Note: The vacuum-handling station is a hypothetical example used to illustrate the engineering and organizational decisions discussed in this article.